1. Does defragmenting improve performance at all on SSD? Could it damage the SSD i.e. wear the SLC out quicker?
At all? Perhaps, but it'd be extrodinarily small.
The reason hard drives slow down when fragmented is that seeking is more expensive than not seeking. When a file is fragmented, the drive must physically move its head (i.e., seek) to each part of the file. Since moving the head takes some ammount of time, any unnescary head movement wastes time. By consolidating the file into one piece the head doesn't need to move, allowing data to be read off as fast as it passes underneath it.
By contrast, there is no seek penalty for solid-state drives. The drives have no moving parts and thus no "head" which must be moved to the data before doing a read. Since no seeking is involved, it takes the same ammount of time to access sector 1000000 if you just asked for sector 1 or sector 999999.
Now, there
might be the tinyest of speed increases reading a defragmented file under certian circumstances...
First, if the SSD does prefetching. That is to say, if the drive decides to grab adjacent sectors (in addition to the one you specifically requested) and store them in a high-speed buffer. A data request "near" to your last one could be satisfied from the buffer instead of the drive, and thus result in faster access. I haven't heard of SSD makers doing this, but its theoretically possible.
Second, if the ATA command set allows you to request multiple sectors with a single command. This would save traffic on the SATA bus, because you could request the entire range of sectors that a file occupies with a single command rather than sending one command per sector. I don't know ATA well enough to say if this is the case. Even if it were though, the tiny command packets would hardly come close to impacting your speed
As for wear, writing to most forms of flash memory will damage them in some way. There are only so many times you can write to an individual block before it goes kaput. Given the imperceptible speed increase that may be had from defragmenting and the wear from all the writing a defrag will cause, I doubt its worth doing.
2. Does short stroking improve the performance at all in transfer rates? I know it will make no difference in seek times on SSD.
Short stroking improves the performance on hard drives for the same reason defragmenting does -- the seek penalty is reduced. If it takes the drive 8ms on average to seek anywhere on the platter, it'll only take 4ms on average to seek anywhere on the first half of the platter, 2ms anywhere in the first quarter, and 1ms anywhere in the first eighth.
Since SSDs don't have a physical head that needs to seek anywhere, short stroking the drive should give you absolutely no benefit. Not even the two possible cases listed above would apply. The first would require that your partitions be no larger than the size of the buffer (which if used would probably not exceed a few MB in size), and partition size has no bearing on the second.
3. What could be done to improve the performance of SSD? I know that getting the specialized MFT firmware upgrade, but it is too expansive for me at the moment since it costs $300.
RAID and overclocking (we're talking about something that probably uses a clock signal internally after all
) are the only things I can think of at the moment. The first has been shown to scale well if you can get a fast enough controller, and the second has never been attempted AFAIK.
JigPu
